A Kalman-smoother-based known route energy demand predictor for road vehicles

Journal article, 2026

The increased demand for electric transportation has brought new challenges, many of which relate to the limited range and long charging time. This may be addressed by providing reliable and accurate residual-range estimates and energy-optimized route selections. Crucially, the performance of such systems is largely determined by the accuracy of the underlying energy demand prediction algorithm. Existing solutions typically predict energy demand but fail to derive the prediction uncertainty, an ever more sought-after quantity. Methods that attempt to provide such a measure often rely on data-driven techniques or computationally intensive Monte Carlo simulations. This paper is therefore set out to provide a dynamic model capable of handling varying operating conditions whilst remaining computationally efficient. By regarding the driver reference speed as a measurement, an observer-like prediction model can be formulated. Building on this principle, a vehicle-independent description is attained, owing to the use of merely exogenous parameters. The observer is realized as a Cubature Rauch-Tung-Striebel Smoother, which estimates both the state and the uncertainty, providing qualitative information about expected variation. The algorithm’s mean prediction is verified by analyzing the velocity profile, the Kalman gain, and the energy demand, all of which exhibit the expected behavior. Its uncertainty estimate is evaluated against real-world vehicle operation data, which show that the considered uncertainty sources significantly contribute to the energy demand prediction uncertainty. Further improvements in accuracy are expected as more uncertainties are accounted for.